Your Best AI Investment | Maybe It's An Old One That You've Already Made

Every year, organisations shell out millions on AI. There are new pilots, some proof-of-concept projects pass the testing phase at the yes-men level, and budgets expand. Senior management continues to seek out that next big thing which will change their business forever – if it only works as promised, or even close to it! It asks the question of a future, which, naturally enough in matters like this one, should also accompany uncertainty.

But lately I began to wonder whether there is something wrong with how organisations frame the problem. Instead of searching for the future AI opportunity, maybe we should look and ask about the past ones first.

And every organisation celebrates its flagship pilots. In turn they become internal success stories, case studies and presentations. So each organisation analyses its failed pilots to understand how things went wrong and what factors led to not getting the desired ROI. Both types of projects are interesting because they appear to teach us some lessons.

But in the past few months, my interest shifted to a different kind of project. The topic of AI poems and what they can bring to active debate is scarcely talked about, analysed or factored into any argument regarding the future of AI.

And these are the half-baked projects that didn't really work or completely flopped. Most organisations have such projects. It had a pilot who seemed somewhat promising from the level of technology but lost that steam once the executive champion changed courses.

A successful proof of concept that solved the problem it set out to solve but was never rolled out because another initiative became a priority.

An AI project delivering impressive results but started without good governance practices, data quality checks and organisational readiness needed for enterprise-scale deployment.

Many years later, they lie buried in the project archives not because somebody has proved them to be worthless but simply because the organisation moved on.

The deeper I looked into examples from digital transformations, the more apparent it became that we need to do something differently with these sorts of projects.

A failed pilot is, as they say, no more. Evidence indicates the proposed solution was impractical, did not deliver sufficient value or was an overall waste of investment.

They are also different from these projects. The one still open is the question business asks.

The technology advances. Capabilities of AI improve. Data gets better. Governance is developed. Regulations become clearer. The organisation gains experience. The new leaders have different priorities. The business problem itself can be changed!

The context at and around the time of that initial pilot may be so strong that there ought to be a re-evaluation about doing this on a larger scale versus doing these pilots again.

One interesting non-AI-related example of this is the following: In the UK, for example, plans to radically redesign and gradually roll out universal credit as part of a digital transformation programme ran into several snags. The main takeaway here is not that a failed programme will turn into a successful one in due course. This is not, therefore, a marker of failure in organisational capability or governance and implementation maturity; rather, it demonstrates just how quickly these capabilities can change over time.

Healthcare is another exciting example we see. Most of the AI initiatives began as narrow pilots and were scaled up in multiple departments once clinical confidence, governance and organisational readiness were ensured. In such cases again, it is not necessarily the technology that has created a change. It was indeed an organisational readiness which arrived gradually over time.

They made me re-evaluate a popular assumption…

The result of a pilot is typically not seen as an absolute verdict.

This must be understood as a judgement on the context.

If this context changes, does the decision need to be changed too?

Just maybe, I am beginning to wonder if the organisations are sitting on this invisible portfolio of unrealised potential.

These became projects I called Cold Pilots™. Not because they failed. Not because they succeeded. Because they faded away quietly, long before their full potential was acknowledged.

The difference may seem subtle, but I think it completely changes the conversation. As long as executives think of projects simply in terms of successes and failures, there will be only one strategic question: What action should we take?

Now, how many of those not-dead-yet AI projects are still working on real business issues?

Organisational readiness issues — these are just a few of the reasons why more pilots never get off (or go on). How many were simply stopped because those in charge realised it would have little value compared to what they had been given?

How many decisions that were correct 3 years ago are still the right ones today?

And even more importantly, how do we know?

Most companies carry out portfolio reviews, and they are doing a great job in evaluating spend, delivery timelines, benefits & risk. However, they are dramatically less useful for assessing whether a changing organisational context has made that project far more (or less) likely to succeed.

As far as I am concerned, this is an emerging executive skill set.

I am working with a framework for this idea based on what we call Recovery Readiness™. It's not about saving all dormant products. Certain projects should be retained in archival status while others are terminated. However, the fun part is identifying which of them have had their chance of success altered by the changing organisations themselves.

Now, this is where AI may truly deliver value in a different light.

Much of what we discussed today is AI generating new recommendations, helping automate or improve decision-making. These applications are very important. I also believe that AI may help organisations to re-remember their created value. By analysing historical project documents, governance records, and lessons learned along with technical challenges and evolving business priorities, AI can identify which dormant projects are worth giving a new lease of life.

One of the best future uses of AI may be helping organisations with enhancements in existing products and processes.

Maybe it reminds them of the value they have already built up.

Most, I suspect, have a portfolio of forgotten pilots. Some belong in history. Some could be worth reviving.

Before paving the way for their next pilot programme, perhaps the board of directors ought to pause and consider a different question.

Where did we tap all that potential value from the AI investments we've already made?

𝑾𝒆𝒍𝒄𝒐𝒎𝒆 𝒕𝒐 𝒕𝒉𝒆 𝒇𝒊𝒓𝒔𝒕 𝒊𝒏 𝒂 𝒔𝒆𝒓𝒊𝒆𝒔 𝒆𝒙𝒑𝒍𝒐𝒓𝒊𝒏𝒈 𝑪𝒐𝒍𝒅 𝑪𝒂𝒔𝒆𝒔™, 𝒂𝒏 𝒆𝒗𝒆𝒓𝒈𝒓𝒆𝒆𝒏 𝒓𝒆𝒔𝒆𝒂𝒓𝒄𝒉 𝒑𝒓𝒐𝒈𝒓𝒂𝒎𝒎𝒆 𝒕𝒉𝒂𝒕 𝒉𝒆𝒍𝒑𝒔 𝒐𝒓𝒈𝒂𝒏𝒊𝒔𝒂𝒕𝒊𝒐𝒏𝒔 𝒊𝒅𝒆𝒏𝒕𝒊𝒇𝒚, 𝒆𝒗𝒂𝒍𝒖𝒂𝒕𝒆 𝒂𝒏𝒅 𝒓𝒆𝒗𝒊𝒕𝒂𝒍𝒊𝒔𝒆 𝒅𝒆𝒇𝒖𝒏𝒄𝒕 𝑨𝑰 & 𝒅𝒊𝒈𝒊𝒕𝒂𝒍 𝒕𝒓𝒂𝒏𝒔𝒇𝒐𝒓𝒎𝒂𝒕𝒊𝒐𝒏 𝒊𝒏𝒊𝒕𝒊𝒂𝒕𝒊𝒗𝒆𝒔 𝒘𝒊𝒕𝒉 𝒍𝒂𝒕𝒆𝒏𝒕 𝒔𝒕𝒓𝒂𝒕𝒆𝒈𝒊𝒄 𝒗𝒂𝒍𝒖𝒆. *it would have little value compared